A slightly faster option:


unsigned char Data[70];
unsigned char CurrentValue;
unsigned char CurrentBitIndex       =3D 0;
unsigned char CurrentByteIndex    =3D 0;

....

if( CurrentBitIndex =3D=3D 0 )
    {
    CurrentValue        =3D Data[CurrentByteIndex];
    if( ++CurrentByteIndex >=3D sizeof Data )
        {
        CurrentByteIndex    =3D 0;
        LoadNewData();
        }
    CurrentBitIndex    =3D 8;
    }

if( CurrentValue & 0x80 )
    SendHiBit();
else
    SendLoBit();

CurrentValue    <<=3D 1;
CurrentBitIndex--;

....


In 7 out of 8 rounds it will skip the first "if" block entirely,
executing just one "AND", one left-shift and one decrement.


Best regards,

Isaac



Em 12/2/2012 16:58, Isaac Marino Bavaresco escreveu:
> Oops! Little mistake, forgot the "( 1 << ( BitNumber & 7 ))".
>
>
> if( Data[ BitNumber >> 3 ] & ( 1 << ( BitNumber & 7 )))
>     SendHiBit();
> else
>     SendLoBit();
> if( ++BitNumber >=3D sizeof Data * 8 )
>     {
>     BitNumber =3D 0;
>     LoadNewData();
>     }
>
>
> Isaac
>
>
> Em 12/2/2012 16:43, Isaac Marino Bavaresco escreveu:
>> Did I understand it right and you shift a chain of 70 data registers and
>> test the lowest bit every round?
>>
>> We discussed many months ago a more efficient way of testing arbitrary
>> bits in an array:
>>
>>
>> char Data[70];
>> int BitNumber =3D 0;
>>
>> ...
>>
>> if( Data[ BitNumber >> 3 ] & ( BitNumber & 7 ))
>>     SendHiBit();
>> else
>>     SendLoBit();
>> if( ++BitNumber >=3D sizeof Data * 8 )
>>     {
>>     BitNumber =3D 0;
>>     LoadNewData();
>>     }
>>
>> ...
>>
>>
>> No real shifting of the data is done, so the size of the array doesn't
>> matter for the efficiency of the process.
>> This routine should be much faster than shifting a chain of 70 bytes.
>>
>>
>> Best regards,
>>
>> Isaac
>>
>>
>>
>>
>>
>> Em 12/2/2012 12:03, PICdude escreveu:
>>> Ah yes, I knew this time would come, where I'd need an efficient way =20
>>> to do efficient bit operations in C, and in this case, shift bits held =
=20
>>> in multiple registers.
>>>
>>> Specifically, I have a stream of ~70 bytes and I need to sequentially =
=20
>>> spit them out an I/O pin, on each timer interrupt (to generate a =20
>>> specific output data stream).
>>>
>>> In assembly, I'd hold these in several registers and do something like.=
...
>>>
>>>        rrf Data9,F
>>>        rrf Data8,F
>>>        ...
>>>        rrf Data0,F
>>>        btfsc STATUS,C      ; C hold current bit value
>>>        goto SendHiBit
>>> SendLoBit:
>>>        ...
>>>
>>>
>>> In C though, it's nice to have single variables that can hold 4 bytes, =
=20
>>> but I can't find a shift operator that let's me extract the last bit, =
=20
>>> so I'm thinking I'd have to so something like ...
>>>
>>> OutBit =3D Data0 & 0b1;      // Extract lowest bit and hold
>>> Data0 >> 1;                  // Assume long (4 bytes)
>>> if (Data1 & 0b1)
>>>        Data0 &=3D 0b1;      // Uppermost bit =3D 1
>>> Data1 >> 1;
>>> if (Data1 & 0b1)
>>>        Data1 &=3D 0b1;
>>> Data1 >> 2;
>>> // Send output bit here...
>>>
>>>
>>> As compact as it looks, it seems so inefficient compared to the =20
>>> assembly version.  (I haven't actually implemented this yet, btw).  Is =
=20
>>> there a better way to do this in C?  I don't actually need to shift if =
=20
>>> there's a more efficient way to read the bit values using some type of =
=20
>>> index variable.
>>>
>>> An incomplete thought running through my head currently is holding "1" =
=20
>>> in some long var, shifting that on each iteration, then using that as =
=20
>>> a mask to "AND" the Data.  I'd also hold another index var (ranging =20
>>> from 0 to 2) which would tell me which data variable I was currently =20
>>> at.  But this method may be even worse.
>>>
>>> Or is this where I do inline assembly?  With inline assembly I'd =20
>>> probably expend several cycles moving the data to know register =20
>>> locations, unless there's some simple way to figure out in RAM where =20
>>> the Data2,1,0 variables are held.
>>>
>>> How would you C pros do it?
>>>
>>> Cheers,
>>> -Neil.

--=20
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist
.